This article was originally published on December 15, 2025 on Law360 and is republished here with permission.

For years, arbitration users have sent a consistent message: They want faster, more predictable, and more cost-effective processes without sacrificing quality and fairness.

The American Arbitration Association–International Centre for Dispute Resolution’s announcement and rollout of an artificial intelligence arbitrator for two-party, documents-only construction disputes this September marks a meaningful response to that market demand.[1]

While the tool is intentionally very narrow at this stage, its launch is significant because it provides a technology-driven, arbitrator-supervised workflow that aims to deliver speed and cost savings in a limited and well-defined segment of smaller construction cases.

The AI arbitrator, however, is less a radical departure than a pragmatic iteration: It attempts to make arbitration work the way users wish it would in a narrow band of disputes. If early results bear out its promise, the implications for U.S. and international arbitration in 2026 and beyond could be felt well beyond its initial application to documents-only construction cases.

Framing the Shift: A Response to Market Demands

The market signals have been consistent for more than a decade across arbitration sectors and jurisdictions. Users prefer arbitration for complex and cross-border disputes, but they also report that inefficiency — driven by adversarial tactics, over-lawyering and insufficiently proactive case management by arbitrators — risks eroding much of its value.[2]

The 2025 Queen Mary University of London International Arbitration Survey brought the issue into focus.[3] Respondents prioritized tools that shorten timelines and lower costs and are receptive to technical innovation, including AI, when it is explainable, disclosed and supervised.

Institutions and seats have acted accordingly. In the past18-24 months, arbitral rules and legislation have been modernized specifically to address speed and cost — e.g., the Hong Kong International Arbitration Centre‘s 2024 rules,[4] the Singapore International Arbitration Centre‘s 2025 rules[5] and legislative refinements in key jurisdictions such as the 2025 U.K. Arbitration Act[6] — all aimed at earlier case management, stronger summary tools, clearer expedited tracks and more efficient routes to enforceable awards.

Against this backdrop, the AAA-ICDR’s AI arbitrator looks less like an experiment and more like the next step in a gradual progression: a deliberately scoped, opt-in feature designed to trim time and money from a class of disputes where records are relatively standardized, and hearings are rare. Put differently, the AI arbitrator is the most readily deployable AI solution for arbitration — designed to answer users’ repeated demands for time and cost savings for a narrow set of disputes that lend themselves to this tool.[7]

What the AI Arbitrator Is — and What it Isn’t

At bottom, the AI arbitrator is not a robot judge. It is a supervised workflow for a very specific subset of cases.

What it is:

• A narrow, opt-in product for two-party, documents-only construction disputes administered by the AAA-ICDR. Parties must affirmatively choose it; otherwise, the case proceeds under traditional procedures.[8]

• A system trained on a large body of AAA documents-only construction awards and was refined with expert-labeled examples. It uses structured prompts and conversational AI to produce a draft award responsive to the parties’ submissions.[9]

• A human-in-the-loop process where every case is overseen by a human arbitrator who reviews, edits, challenges, and ultimately signs and issues the final award. The human remains responsible for the decision and its reasoning.[10]

• A transparency-focused tool that seeks to afford users a greater understanding of how their confidential information will be used and protected. The AAA-ICDR has published FAQs, governance principles and explanatory materials addressing the tool’s scope, training data, ethics and privacy.[11] The platform’s value proposition is expressly tied to explainability and accountability.[12]

• A time- and cost-saver for a specific subset of construction arbitrations. The AAA-ICDR claims that the AI arbitrator tool will generate expected time savings starting around 20%-25% and cost reductions starting around 35% in appropriate documents-only construction disputes — results that, if sustained, would directly answer user priorities.[13]

What it isn’t:

• It is not an unsupervised or autonomous decision-maker. The AI produces a draft; a human arbitrator decides. There is no award without human judgment and signature.

• It is not a replacement for evidentiary hearings or complex case management. The initial deployment excludes multiparty cases, oral hearings and complex factual matrices. The tool is tailored to the repeatable patterns and record-based nature of documents-only construction disputes.

• It is not a black box. The program is framed around disclosure and governance: The institution has surfaced FAQs, ethics and privacy standards, and dedicated content about human oversight, with pathways for case-by-case clarification.

• It is not an all-purpose cure for arbitration’s challenges. The tool does not eliminate adversarial behavior, solve discovery excesses or rewrite parties’ incentives. It is a targeted intervention where the data and the procedure are conducive to reliable automation with human oversight and validation.

Why the Narrowness Matters

The narrowness of the AAA-ICDR’s AI arbitrator tool is what makes this initiative particularly apt for the subject matter disputes it is intended to manage. Documents-only construction disputes are typically low dollar and relatively standardized, and often involve well-developed contract frameworks that lend themselves to structured analysis.

Maybe more importantly, they are also the very disputes for which users most often demand speed, clarity and predictable cost. Indeed, within the construction sector itself, low-value, document intensive claims are commonplace, and the time and expense required to prosecute those claims in an arbitration can often be cost-prohibitive.

By focusing there, the AAA-ICDR can measure results, refine processes and build trust under conditions where human-in-the-loop oversight can be most effective and transparent. If the institution can consistently deliver faster timelines and cost savings in this area, and if users report comfort with quality and fairness, the market case for cautiously expanding the tool set becomes straightforward.[14]

What This Means for the Future — United States and Worldwide

In the U.S. and abroad, several near-term shifts are likely.

First, we expect the AI arbitrator tool to normalize AI-aware procedural discussions.

Parties and tribunals already address document formats, confidentiality protocols and discovery parameters at the first procedural conference; now they may contemplate adopting the use of AI to these procedural steps in the case. Counsel will ask institutions for their AI policies and governance, and parties may negotiate — or at least reserve — opt-in/opt-out choices where an AI-assisted path is available.

Sophisticated users, particularly in sectors with repeatable disputes, such as construction, supply chain and tech procurement, may even begin to run pilots and track metrics like cycle time, fee spend, award quality and postaward challenges. If those metrics meet expectations, there is a greater possibility for the expansion of similar AI arbitrator tools into other sectors and disputes.[15]

Second, clause drafting will evolve.[16] In transactions where documents-only arbitration is already common, parties are likely to experiment with AI-enabled clause variants that (1) specify eligibility criteria (e.g., two-party, below a stated amount); (2) preserve the right to opt-in or opt-out at filing; (3) incorporate disclosure baselines for AI use by counsel, experts and the tribunal; and (4) align fees and timetables to expedited tracks.

Because the AAA-ICDR’s approach is opt-in and human-supervised, arbitration agreement drafters may have the ability (and inclination) to test how best to adopt AI arbitrator tools without foreclosing traditional processes.

Third, early skepticism by local courts is likely inevitable but the involvement of human arbitrators and other safeguards implemented by institutions like the AAA-ICDR should give jurisdictions comfort that the process is consistent with prevailing award enforcement standards. Because a human arbitrator reviews, edits and issues the award, it seems more likely that award challenges are likely to focus on disclosure, due process in the conduct of proceedings, and whether the human decision-maker exercised independent judgment.

These are matters that the courts already assess when they hear allegations of evident partiality or procedural irregularity. The more transparent the process — and the clearer the record of human oversight — the more comfortable courts are likely to be with enforcement.

Fourth, arbitrator competencies will likely shift at the margins. As tribunals grow comfortable using AI responsibly for discrete tasks, e.g., issue-list generation, draft preparation and citation checks, institutional guidance and community norms will coalesce around disclosure and quality-control standards. That trend is already visible in emerging institutional guidelines and commentary. Arbitrators who can operate confidently within those guardrails — maintaining confidentiality, avoiding overreliance and documenting validation — may be best placed to address the demands an evolving practice.

Internationally, this step will add fuel to the ongoing competition among institutions and seats.

Major institutions have spent the past few years enhancing summary tools, tightening expedited tracks and modernizing case management. The AAA-ICDR’s move introduces a tangible benchmark for AI-enabled administration that others can watch, imitate or seek to improve.

The current trends among seats and institutions suggest that we may see (1) additional AI arbitrator pilots where records are standardized and hearing time is rare, (2) harmonization of AI disclosure protocols in institutional rules and practice notes, and (3) more granular data-sharing about speed, cost and user satisfaction. In that environment, the institutions that credibly quantify improvements may very well retain a competitive advantage among users.

Overall, arbitration user demands and expectations for efficiency and timeliness are likely to become sharper. For repeat players, AI-assisted arbitration will be evaluated like any process: run a controlled trial; measure cycle time and cost variance; assess quality (including the rate and outcome of any post-award challenges); and scale if performance meets expectations.

This practical, data-driven approach — rather than ideology for or against AI — is most likely to determine the pace and extent of adoption.

An Inflection Point?

In many ways, 2025 was the year AI shifted in arbitration from promise to pilot.

The AAA-ICDR’s AI arbitrator is not a silver bullet, and it is not meant to be. Its relevance lies in how directly it confronts what users have been telling the market for years: Make arbitration faster and more cost-effective in the kinds of cases that most need it, and do so with transparency and accountability. By choosing a narrow scope, keeping human decision-makers responsible and publishing governance materials, the AAA-ICDR has set a pragmatic template that others may scrutinize, adapt and scale.

The coming years will bring more data and more sophisticated generative artificial intelligence. If users report that the process is faster, cheaper and comparably fair — and courts are comfortable with the supervision and disclosure that surround it — expect measured expansion in the U.S. and carefully tailored pilots abroad.

If the results disappoint, the market will adjust. Either way, the signal from 2025 is unmistakable: The use of generative AI in arbitration is moving from conversation to implementation, and the user’s definition of value — speed, cost and trust — now sits at the center of that effort.

---

Albert Bates and Zach Torres-Fowler are partners at Troutman Pepper Locke LLP.

The opinions expressed are those of the author(s) and do not necessarily reflect the views of their employer, its clients, or Portfolio Media Inc., or any of its or their respective affiliates. This article is for general information purposes and is not intended to be and should not be taken as legal advice.

[1] AAA-ICDR, AAA-ICDR to Launch AI-Native Arbitrator, Transforming Dispute Resolution, dated September 17, 2025, available at https://www.adr.org/press-releases/aaa-icdr-to-launch-ai-native-arbitrator-transforming-dispute-resolution/.

[2] Queen Mary University of London & White & Case, 2025 International Arbitration Survey – The path forward: Realities and opportunities in arbitration (2025).

[3] Queen Mary University of London & White & Case, 2025 International Arbitration Survey – The path forward: Realities and opportunities in arbitration (2025).

[4] Hong Kong International Arbitration Centre, 2024 Administered Arbitration Rules in effect as of 1 June 2024.

[5] Singapore International Arbitration Centre, Arbitration Rules of the Singapore International Arbitration Centre, 7th Ed., in effect as of 1 January 2025.

[6] Arbitration Act 2025, 2025 c. 4 (UK).

[7] AAA-ICDR, AAA-ICDR AI Arbitrator Now Available for Documents-Only Construction Cases, dated Nov. 3, 2025 available at https://www.adr.org/press-releases/aaa-icdr-ai-arbitrator-now-available/; see also AAA-ICDR, AAA-ICDR InnovAAAtion hub, available at https://www.adr.org/innovaaation/.

[8] AAA-ICDR, AAA-ICDR AI Arbitrator Now Available for Documents-Only Construction Cases, dated Nov. 3, 2025 available at https://www.adr.org/press-releases/aaa-icdr-ai-arbitrator-now-available/; AAA-ICDR, AI Arbitrator: A New Path to Dispute Resolution FAQs (2025) available at https://www.adr.org/media/1vva0v2y/ai-arbitrator-one-sheet-20251029-1.pdf.

[9] AAA-ICDR, AAA-ICDR AI Arbitrator Now Available for Documents-Only Construction Cases, dated Nov. 3, 2025 available at https://www.adr.org/press-releases/aaa-icdr-ai-arbitrator-now-available/; AAA-ICDR, AI Arbitrator: A New Path to Dispute Resolution FAQs (2025) available at https://www.adr.org/media/1vva0v2y/ai-arbitrator-one-sheet-20251029-1.pdf.

[10] AAA-ICDR, How the AAA’s AI Arbitrator Keeps Humans in the Loop available at https://www.adr.org/news-and-insights/ai-arbitrator-faq/; AAA-ICDR, AI Arbitrator: A New Path to Dispute Resolution FAQs (2025) available at https://www.adr.org/media/1vva0v2y/ai-arbitrator-one-sheet-20251029-1.pdf.

[11] AAA-ICDR, AI Arbitrator: A New Path to Dispute Resolution FAQs (2025) available at https://www.adr.org/media/1vva0v2y/ai-arbitrator-one-sheet-20251029-1.pdf.

[12] AAA-ICDR, AAAi Standards for AI in ADR, available at https://www.adr.org/media/pgfe4xwg/aaai_standards_for_ai_in_adr.pdf.

[13] AAA-ICDR, AI Arbitrator: A New Path to Dispute Resolution FAQs (2025) available at https://www.adr.org/media/1vva0v2y/ai-arbitrator-one-sheet-20251029-1.pdf.

[14] AAA-ICDR, AI Arbitrator: A New Path to Dispute Resolution FAQs (2025) available at https://www.adr.org/media/1vva0v2y/ai-arbitrator-one-sheet-20251029-1.pdf.

[15] As noted above, AI Arbitrator was trained on a database of approximately 1400 AAA documents-only construction cases with awards and was refined with expert-labeled examples. This is a relatively unique database of cases that proceeded on documents only, i.e., without fact or expert witness testimony. One of the challenges to the expanded use of such a tool is identification of other caseloads with circumscribed datasets that can be used to properly develop and train the AI program.  

[16] The AAA has also launched ClauseBuilder AI (Beta), a generative AI tool designed to simplify drafting arbitration and mediation clauses.  AAA, ClauseBuilder AI & API Tools Streamline Arbitration Drafting (2024) available at https://www.adr.org/news-and-insights/introducing-clausebuilder-ai-beta-and-api-innovations-streamlining-arbitration-and-mediation-clause-drafting-with-generative-ai/.